Multi-hop Question Answering on 2WikiMultiHopQA (LM/F1/Acc)
70.11LM ScoreGoldRouter (Multi-RAG-Agent)
Evaluation Results
| Method | Links | |||
|---|---|---|---|---|
| GoldRouter (Multi-RAG-Agent)Models=Qwen-Plus2025.01 | 70.11 | 72.37 | 76.33 | |
| EfficientRAG (Single-RAG-Agent)Models=Qwen-Plus2025.01 | 69.41 | 71.77 | 75.85 | |
| GoldRouter (Multi-RAG-Agent)Models=LLaMA-3.1-8B2025.01 | 49.73 | 52.99 | 52.07 | |
| EfficientRAG (Single-RAG-Agent)Models=LLaMA-3.1-8B2025.01 | 48.45 | 52.04 | 51.13 | |
| CoT (Without RAG)Models=Qwen-Plus2025.01 | 42.38 | 44.89 | 37.31 | |
| CoT (Without RAG)Models=LLaMA-3.1-8B2025.01 | 31.4 | 34.46 | 25.99 |